XPQRS: Expert power quality recognition system for sensitive load applications

• Published in: Measurement : journal of the International Measurement Confederation 2023-07, Vol.216, p.112889, Article 112889
• Main Authors: Khan, Muhammad Umar, Aziz, Sumair, Usman, Adil
• Format: Article
• Language: English
• Subjects: Approximated derivatives; Feature extraction; Power quality disturbances; Single power cycle; Support vector machine
• ISSN: 0263-2241; 1873-412X
• DOI: 10.1016/j.measurement.2023.112889

Power Quality (PQ) disturbances of a few milliseconds duration may lead to malfunctioning the sensitive devices connected to the network. This necessitates detecting these PQ disturbances in a minimum time before affecting the sensitive load and adopting countermeasures. An expert PQ recognition system (XPQRS) is proposed to identify seventeen types of PQ events using only one power cycle acquired at 5 kHz. In XPQRS, the input PQ signal is first concatenated with its four derivatives to highlight the deviations. Log energy, Shannon energy, and mobility parameters, extracted from the resultant vectors, are then fed to Quadratic Support Vector Machine. XPQRS offers 96.5% accuracy and provides fast decisions in 34.99 ms which makes it suitable for its implementation on an embedded system. XPQRS shows noise immunity and outperforms the previous studies in terms of decision time, accuracy, and computational complexity. •XPQRS is proposed for sensitive load applications to classify 17 types of PQDs.•The first study to use only one cycle of power signal acquired at 5 kHz.•Derivative-based operators are used to magnify the deviations in the input PQDs.•XPQRS achieved a mean accuracy of 96.5% while performing fast decisions in 34.99 ms.•XPQRS shows noise immunity and good performance as compared to previous studies.